New ML framework predicts crystal structures from electron diffraction data

Researchers have developed ED-CSP, a novel machine learning framework designed to predict crystal structures from electron diffraction data. This framework combines a relational set encoder, a permutation-invariant aggregation method, and a periodic flow generator to accurately determine lattice parameters and atomic coordinates. Trained on a newly constructed dataset called ED-CS, comprising 4.85 million simulated crystal structures, ED-CSP demonstrated strong performance on held-out data, outperforming existing state-of-the-art models. AI

IMPACT Establishes a new benchmark for generative crystal structure prediction, potentially accelerating materials science research.

RANK_REASON The cluster describes a new machine learning framework and dataset for a scientific prediction task, published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Arnaud Demortière
• CHILI-100K
• ED-CS
• ED-CSP
• PXRDGen

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →